In the UK, stroke is the most common serious neurological disease (6) and a leading cause of death (7); there are more than 1.3 million stroke survivors (8), of whom more than 50% have a disability (6). Improving outcomes from stroke is thus a key healthcare priority with a focus on how innovation can facilitate change.
About 80% of acute strokes are ischaemic (9), with large vessel occlusions (LVOs) accounting for around 24 – 46% of all ischaemic strokes (10). Early treatment is critical to rescue potentially salvageable tissue (11) and until relatively recently, the only licensed treatment for acute ischaemic stroke was intravenous thrombolysis (IVT) with recombinant tissue-type plasminogen activator (IV r-tPA). However, since 2014, mechanical thrombectomy (MT) has revolutionised the care of patients with acute ischaemic stroke due to LVO. The efficacy of this treatment is unmatched by any previous therapy in stroke medicine and, when administered within six hours of onset of symptoms, it can reduce brain damage and prevent or limit long-term disability (12).
Identification of patients for MT requires specialist radiological image interpretation in real-time, which is not readily available in most hospitals 24/7 and 365 days of the year. If patients are found to have had an LVO stroke that is appropriate for MT, they are taken by ambulance to a comprehensive stroke centre (CSC) where the procedure can take place. In some cases, the patient is received by the acute stroke centre (ASC) first, which includes several additional steps, such as the patient being administered IVT before being transferred by ambulance to the CSC. Achieving a ‘door-in-door-out’ (DIDO) time of less than 60 minutes at an ASC ensures that patients have similar outcomes to those arriving directly at a CSC (1). Nagaratnam et al 2021 has shown that implementing AI into the standard stroke pathway at an ASC can significantly reduce DIDO times (2).
The NHS Long Term Plan in 2019 (3) set out national targets for stroke care including increasing rates of MT from 1% to 10% by 2022 and rates of IVT from 11% to 20% as both treatments can significantly reduce the severity of disability caused by a stroke. Our initial analysis has shown that whilst rates of MT vary across evaluation sites, they continue to increase despite the impact of Covid-19. The evaluation average has risen from 1.7% to 3% from January 2019 to March 2022 with some evaluation sites achieving rates in excess of 7%. We also know from our qualitative findings that more than half of people asked believed that the use of e-Stroke leads to the identification of more eligible patients for MT.
In contrast, and in part mirroring our findings through our evidence review, rates of thrombolysis appear to be decreasing on average from 13.4% to 11% retrospectively, with a few sites remaining stable or showing an increase. However, the rates again do seem to be recovering from a post-Covid low of 9.3%. We know from Allen et al (4) that five out of ten patients who were treatable but did not receive IVT were because doctors chose not to proceed with treatment when other higher-thrombolysing hospitals would have done. From our qualitative analysis, it would appear that e-Stroke does not have an impact on the decision time to administer IVT. Further analysis will investigate whether more eligible patients could benefit from IVT and MT through the inclusion of e-Stroke in the image-based treatment pathway, by looking at data before and after e-Stroke implementation.
A crucial part of improving clinical outcomes for patients with an LVO is ensuring that all eligible patients can have a MT. Early identification and quicker decision times, supported by AI can increase the window of opportunity. Our initial analysis shows that, in general, time from first imaging to IVT and MT is, as an evaluation average, increasing. However, we also know from the GIRFT Stroke National Specialty Report (5) that other factors such as 24/7 access to a CSC, availability of staff (particularly INRs) and distance to the CSC, including ambulance transfer and availability, can all negatively impact the time to treatment, and we will consider these factors in our analysis going forward.
Through our qualitative analysis, we found that whilst there was difference in AI and clinical decision, e-Stroke is considered by its users to be decision support software that adds value to the standard of care, despite 60% of clinicians responding with “yes” to having concern with accuracy. Indeed, it is also important to note that despite concerns, e-Stroke is being used for more cases and 83% of survey respondents believe that the software has reduced the time taken to treat patients with MT.
Through statistical analysis, key trends have been identified in terms of the association between certain demographic factors and performance measures. This analysis has been undertaken using pre-AI implementation data to provide a baseline profiling matrix. Our initial findings suggest that:
- the age of a patient has a correlation with the length of stay on a stroke ward
- older and less deprived areas seem to have better access to IVT
- age is associated with a patient requiring paid carer support post stroke.
Further work is being undertaken to assess ASCs and CSCs and determine other mitigating factors.
- Holodinsky JK, Patel AB, Thornton J, Kamal N, Jewett LR, Kelly PJ, Murphy S, Collins R, Walsh T, Cronin S, Power S, Brennan P, O’hare A, McCabe DJ, Moynihan B, Looby S, Wyse G, McCormack J, Marsden P, Harbison J, Hill MD, Williams D. Drip and ship versus direct to endovascular thrombectomy: The impact of treatment times on transport decision-making. Eur Stroke J. 2018 Jun;3(2):126-135. doi: 10.1177/2396987318759362. Epub 2018 Feb 14. PMID: 31008345; PMCID: PMC6460407.
- Nagaratnam, K et al, (2021) New Study Finds e-Stroke Shortens Treatment Times, Helping More Stroke Patients Achieve Functional Independence (prnewswire.co.uk) (last access 25th November 2022)
- NHS Long Term Plan v1.2 August 2019 (last access 25th November 2022)
- Allen M. et al. (2022) Use of Clinical Pathway Stimulation and Machine Learning to Identify Key Lever for Maximising the Benefit of Intravenous Thrombolysis in Acute Stroke. Original published 15 Jul 2022 https://doi.org/10.1161/STROKEAHA.121.038454 (last accessed 25th November 2022)
- Stroke, Getting it Right First Time; GIRFT Programme National Specialty Report April 2022
- Adamson, J., Beswick, A. and Ebrahim, S. (2004) ‘Is stroke the most common cause of disability?’, Journal of Stroke and Cerebrovascular Diseases, 13(4). doi: 10.1016/j.jstrokecerebrovasdis.2004.06.003.
- NICE (2016) Mechanical clot retrieval for treating acute ischaemic stroke. Available at: https://www.nice.org.uk/guidance/ipg548. (last accessed 28/09/2022)
- Stroke Association Stroke statistics. Available at: https://www.stroke.org.uk/what-is-stroke/stroke-statistics “There are 1.3 million stroke survivors in the UK” (last accessed 28/09/2022)
- Luengo-Fernandez, R. et al. (2013) ‘Quality of life after TIA and stroke: Ten-year results of the oxford vascular study’, Neurology, 81(18). doi: 10.1212/WNL.0b013e3182a9f45f. https://n.neurology.org/content/81/18/1588 (last accessed 28/09/2022)
- Rennert RC, Wali AR, Steinberg JA, Santiago-Dieppa DR, Olson SE, Pannell JS, Khalessi AA. Epidemiology, Natural History, and Clinical Presentation of Large Vessel Ischemic Stroke. Neurosurgery. 2019 Jul 1;85(suppl_1):S4-S8. doi: 10.1093/neuros/nyz042. PMID: 31197329; PMCID: PMC6584910. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6584910/#:~:text=Large%20vessel%20occlusions%20(LVOs)%2C,46%25%20of%20acute%20ischemic%20strokes (last accessed 28/09/2022)
- Saver JL. Time is brain—quantified. Stroke2006;37:263–6. 10.1161/01.STR.0000196957.55928.ab https://pubmed.ncbi.nlm.nih.gov/16339467/ (last accessed 28/09/2022)
- Evans MRB, White P, Cowley P, Werring DJ. Revolution in acute ischaemic stroke care: a practical guide to mechanical thrombectomy. Pract Neurol. 2017 Aug;17(4):252-265. doi: 10.1136/practneurol-2017-001685. Epub 2017 Jun 24. PMID: 28647705; PMCID: PMC5537551. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5537551/ (last accessed 28/09/2022)